New Work Demands New Training

Professional education was designed to produce people who could do the work—execute tasks competently, navigate structures, coordinate with colleagues. For decades, this worked just fine.

But the era of AI doesn't need people who can do the work. It needs people who can get work done—a phrase that sounds similar but describes an entirely different capability. Think of it as a shift from syntax to semantics: the grammar of business tasks becomes something AI handles. What remains stubbornly human is meaning: what are we actually trying to accomplish, and is this output accomplishing it?

This shift is playing out across several dimensions, changing what we need to learn, and thus also what to teach.

From Individual-in-Organization to Individual-as-Organization

You were trained to fit into systems—to play your position well and trust the organization to handle the rest. That made sense. Complex work needed specialized roles, coordination mechanisms, institutional infrastructure. No individual could do it alone.

That constraint is dissolving. With AI, you can individually orchestrate work at scales that used to require teams. But this means taking on responsibilities you were likely never explicitly taught: deciding what to do versus delegate, managing the work of others, all while ensuring quality output.

A marketing manager who used to coordinate with designers, copywriters, and analysts now orchestrates AI for each function—sequencing workflows, checking quality along the process, deciding when to step in and manually make changes. She's not joining an organization anymore. She's becoming one.

This is what it means to be a "firm of one": not a freelancer (that's just a small-scale executor), but someone managing AI-augmented processes the way a CEO manages business units. The grammar of coordination—task handoffs, role boundaries—becomes AI's domain. You provide the meaning: what is this whole operation actually for?

From Execution to Process Design

Out of that scale, another challenge arises. You were trained to execute: learn methods, apply them reliably, produce outputs. But if a best practice can be fully specified, it becomes the AI's job. The value of knowing "the right way to do X" plummets.

This doesn't make process knowledge irrelevant—it changes its function. Instead of executing processes, you design them. Instead of following best practices, you define what "best" means here. Instead of handling every case, you handle exceptions—situations outside what the process was designed for.

A financial analyst who used to build DCF models now defines what "conservative assumptions" means for this specific deal—what could go wrong and how to model it. That part is complex and situational; how to instantiate in a spreadsheet is a known best practice—grammar rather than meaning.

The deeper skill isn't tool proficiency. It's knowing what should be systematized versus what requires case-by-case discernment—and recognizing when a situation has crossed from one to the other.

From Interpersonal to Intrapersonal

Here's where you hit the deepest layer. You can't design good processes—or direct a firm of one—without knowing what you actually want. And by scaling up your stated goals, AI has a way of revealing that you often don't really know what you want.

Professional education devotes enormous energy to interpersonal skills: leadership, negotiation, team dynamics. That complexity isn't going away; human-to-human coordination remains essential. But AI adds something new. AI systems are powerful mirrors. They surface your confusion, unstated assumptions, and fuzzy thinking simply by doing what you asked. If you don't know what you want, AI will cheerfully produce exactly what you asked for—revealing the gap between what you said and what you meant.

A designer keeps rejecting AI designs for a logo as "not quite right" but can't explain why. The AI has mastered the grammar of her requests—they are all indeed visually appealing logos. What's missing is meaning he hasn't articulated—criteria he's never made explicit, even to himself.

This makes self-knowledge a professional skill. Introspection, clarity of intent, awareness of your own patterns—these move from "nice to have" to core competency. The professional who succeeds with AI won't just be good at reading other people. They'll be good at reading themselves.

What This Means

These three shifts share a structure: AI absorbs the grammar of work, concentrating human value in the meaning. Direction, discernment, judgment. Knowing what you want. Knowing if you got it. Knowing when the rules don't apply.

Here's a simple test for any skill you're developing or teaching: Could this be fully specified as instructions to an AI? If yes, its value is declining. If no—if it requires the kind of contextual judgment, self-knowledge, or exception-handling that resists specification—its value is (relatively) increasing.

Students should weight their learning toward the unspecifiable: defining what "good" means in novel situations, recognizing when standard approaches fail, articulating intent clearly enough to direct others (human or AI). This isn't easy as these capabilities are less like classroom skills and more like behaviors and practices. In response, teachers can help by rethinking lessons and projects around these capacities—less "execute this process correctly," more "define what success looks like here and evaluate whether we achieved it."

The skills that survive are the ones that can't be written down.